WO2017206099A1 - Method and device for image pattern matching - Google Patents

Abstract

Provided is a method for image pattern matching for reducing the duration of image matching in a pattern matching algorithm. The method comprises: obtaining an original template image and an original target image; creating k+1 layers of template image pyramid and target image pyramid; obtaining a normalized cross-correlation graph with a normalized cross-correlation value greater than a preset threshold and a corresponding angle graph, wherein the pattern matching between the topmost template image and the topmost target image is a full angle pattern matching, and the pattern matching between the template images of the other layers and the target images of corresponding layers is interval angle pattern matching; and in the first layer, obtaining a normalized cross-correlation graph C₁ and a corresponding angle graph A₁, wherein the area corresponding to the C₁ represents a position where the original template image appears in the original target image, and the corresponding angle graph A₁ is a rotation angle corresponding to the original template image.

Description

Method and device for image pattern matching Technical field

The invention belongs to the technical field of image processing and the field of computer technology, and in particular relates to a method and device for image pattern matching.

Background technique

Image matching technology is an important research in the field of digital image processing and has been widely used in computer vision, virtual reality scene generation, aerospace remote sensing measurement, medical image analysis, optical and radar tracking, and landscape guidance.

Generally speaking, since the imaging conditions obtained by different images at different times, different sensors, and different viewing angles are different, even the geometrical characteristics, optical characteristics, and spatial positions exhibited in the image may be greatly different for the same object. If the effects of noise, interference, etc. are taken into account, the image will be greatly different. Image matching is to find the same point through these differences. Image matching algorithms are mainly divided into two categories: one is based on grayscale matching; the other is based on feature matching.

According to the known pattern (template map), the process of searching for a matching sub-image into another map is called pattern matching. Pattern matching is an important algorithm in machine vision; gray value pattern matching is one of the earliest and most widely used algorithms. Gray value pattern matching generally uses Normalized Cross Correlation (NCC) between the template image and the target image area as a measure similarity criterion. Pattern matching in a more generalized manner includes searching for a template image in which a rotation transform and a scale transform are performed in the target image.

Most machine vision software currently implements a gray-scale pattern matching module based on normalized cross-correlation, and supports full-angle search and limited zoom range search. For example, Cognex's In-Sight Explorer supports a maximum 360 degree rotation angle and a 10% scaled gray value pattern match. In the prior art, pattern matching of full-angle search is used in both high-resolution images and low-resolution images, which increases the computational complexity of pattern matching.

Summary of the invention

Embodiments of the present invention provide a method and apparatus for image pattern matching, which are used to improve pattern matching. The rate at which images are matched in the algorithm.

In view of this, the first aspect of the present invention provides a method for image pattern matching, which may include:

Acquiring the original template image with the original target image;

Establishing a k+1 layer template image pyramid and a target image pyramid, the template image pyramid is reduced from the lowest layer to the top layer, and the area is reduced by layer by layer, and the bottom layer area of the template image pyramid is the area of the original template image, The target image pyramid is reduced from the lowest layer to the top layer, and the area is scaled down by layer. The bottom layer area of the target image pyramid is the area of the original target image, and k is an integer greater than or equal to 1;

Performing pattern matching on the template image of each layer and the target image of the corresponding layer to obtain a normalized cross-correlation graph C and a corresponding angle map A with normalized cross-correlation values greater than a preset threshold; wherein, the topmost template image The pattern matching with the topmost target image is a full-angle pattern matching, and the pattern images of the other layer's template image and the corresponding layer's target image are matched to the interval angle pattern matching; the topmost target image is reduced by the original target image. At the top level, the target images of the other layers are all sampled from the normalized cross-correlation map acquired by the pattern matching to the next layer.

The first layer 1, to obtain a normalized cross-correlation is normalized cross-correlation value is greater than the predetermined threshold value is a C ₁ and an angle corresponding to FIG. A _1, the C ₁ corresponding to an area represented by the original template image in the original A position appearing in the target image, the angle corresponding to the corresponding angle map A ₁ indicating the rotation angle of the original template image at the position where the original target image appears.

With reference to the first aspect of the embodiments of the present invention, in a first possible implementation manner of the first aspect of the embodiment of the present invention, the matching of the top-most template image with the top-level target image may include: Performing a full-angle pattern matching between the template image of the k+1th layer and the target image of the k+1th layer, and obtaining a normalized cross-correlation graph _Ck+1 with a normalized cross-correlation value greater than the preset threshold and corresponding FIG angle a _{k + 1,} and the normalized cross-correlation C _{k + 1} and FIG angle corresponding to FIG sample a _{k + 1} through k-th layer is a layer the target image C k _'k and the corresponding angle a in FIG. ' _k .

With reference to the first possible implementation manner of the first aspect of the embodiment of the present invention, in a second possible implementation manner of the first aspect of the embodiment, the full angle is [-π, π],

The matching of the template image of the k+1th layer with the target image of the k+1th layer may include: using the k+ in the full angle [-π, π] according to the step size x _k+1 The template image of the first layer performs pattern matching on the target image of the k+1th layer, and the _{k k+1} is a positive integer.

With reference to the first aspect of the embodiments of the present invention, in a third possible implementation manner of the first aspect of the embodiment of the present invention, the interval angle pattern matching between the template image of each layer and the target image of the corresponding layer may include : matching the template image of the layer a and the target image C′ _a of the layer a to the interval angle pattern of the layer a, and obtaining a normalized cross-correlation graph _Ca and a normalized cross-correlation value greater than the preset threshold. Corresponding angle map A _a , and up-sampling the normalized cross-correlation map _Ca and the corresponding angle map A _a to the target image C' _a-1 of the a-1 layer and the corresponding angle Figure A' _a-1 , 1 < a ≤ k, and a is a positive integer.

With reference to the third possible implementation manner of the first aspect of the embodiment of the present invention, in a fourth possible implementation manner of the first aspect of the embodiments, the method may further include: determining the target of the other each layer In each connected subset of the image C' _a , the position of the normalized cross-correlation value corresponding to the angle on A' _a is

The

According to the

Determine the interval angle of the a layer

(n _a is a positive integer).

With reference to the third possible implementation manner of the first aspect of the embodiment of the present invention, the fourth possible implementation manner of the first aspect of the embodiment of the present invention, the fifth possible In an implementation manner, the matching the template image of the layer a and the target image C′ _a of the layer a to the interval angle pattern of the layer a may include:

According to the steps of using the first layer a x _a template image of a target image layer C _'a pattern matching, the x _a positive integer.

With reference to the second possible implementation manner of the first aspect of the embodiment of the present invention, in a sixth possible implementation manner of the first aspect of the embodiment of the present invention, when k<3, x _k+1 =2 ^k Performing pattern matching on the target image of the k+1th layer using the template image of the k+1th layer according to the step size x _k+1 in the full angle [-π, π] may include: at the full angle [- π, π] of the target image k + 1 th layer using a template image for pattern matching the first k + 1 2 ^k layer according to the step.

A second aspect of the embodiments of the present invention provides an apparatus for image pattern matching, including:

a first acquiring module, configured to acquire an original template image and an original target image;

A module is created for establishing a template image pyramid and a target image pyramid of the k+1 layer. The original template image pyramid is reduced from the lowest layer to the top layer, and the area is reduced by layer by layer. The bottom layer area of the template image pyramid is the original The area of the template image, the target image pyramid is reduced from the lowest layer to the top layer, and the area is scaled down by layer. The bottom layer area of the target image pyramid is the area of the original target image, and the k is an integer greater than or equal to 1;

The pattern matching module is configured to pattern match the template image of each layer with the target image of the corresponding layer, and obtain a normalized cross-correlation graph C and a corresponding angle map with the normalized cross-correlation value greater than a preset threshold. A; wherein the pattern of the topmost template image matches the pattern of the topmost target image is a full angle pattern matching, and the pattern image of the other layer template image matches the target image of the corresponding layer is an interval angle pattern matching; the topmost target image For the original target image to be reduced to the topmost level, the target images of the other layers are all sampled by the previous layer for pattern matching, and the normalized cross-correlation map is sampled to the next layer;

a second obtaining module, when used in the first layer, obtains a normalized cross-correlation graph C ₁ and a corresponding angle graph A ₁ whose normalized cross-correlation value is greater than the preset threshold, and the region corresponding to the C ₁ represents the original the template appears in the position of the original image in the target image, a ₁ corresponding to an angle corresponding to the angle of the rotation angle view showing the position of the original template image appearing in the original target image.

With reference to the second aspect of the embodiments of the present invention, in a first possible implementation manner of the second aspect of the embodiments of the present invention,

The pattern matching module is configured to perform a full-angle pattern matching between the template image of the k+1th layer and the target image of the k+1th layer, and obtain a normalized cross-correlation whose normalized cross-correlation value is greater than the preset threshold. Figure C _k+1 and the corresponding angle map A _k+1 , and upsample the normalized cross-correlation map C _k+1 and the corresponding angle map A _k+1 to the target image C of the kth layer ' _k and the corresponding angle map A' _k .

With reference to the first possible implementation manner of the second aspect of the embodiment of the present invention, in a second possible implementation manner of the second aspect of the embodiment, the full angle is [-π, π],

The pattern matching module is further configured to perform pattern matching on the target image of the k+1th layer by using the template image of the k+1th layer according to the step size _xk+1 in the full angle [-π, π], x _k+1 is a positive integer.

With reference to the second aspect of the embodiments of the present invention, in a third possible implementation manner of the second aspect of the embodiment of the present invention, the pattern matching module is further configured to use the template image of the layer a and the layer a The target image C' _a performs interval angle pattern matching of the a layer, obtains a normalized cross-correlation map _Ca and a corresponding angle map A _a whose normalized cross-correlation value is greater than the preset threshold, and normalizes the image The cross-correlation map C _a and the corresponding angle map A _a are upsampled to the target image C' _a-1 of the a-1 layer and the corresponding angle map A' _a-1 , 1 < a ≤ k, And a is a positive integer.

With reference to the third possible implementation manner of the second aspect of the embodiment of the present invention, in a fourth possible implementation manner of the second aspect of the embodiment, the device further includes:

A first determination means for determining 'of each of _a subset of the communication, the maximum normalized cross-correlation values corresponding to the position A' of the target image on each other on the angle of _a C, to

The

a second determining module for

Determine the interval angle of the a layer

(n _a is a positive integer).

With reference to the third possible implementation manner of the second aspect of the embodiment of the present invention, the fourth possible implementation manner of the second aspect of the present invention, the fifth possible implementation manner of the second aspect of the embodiment of the present disclosure In the mode matching module, specifically used in the interval angle

According to the steps of using the first layer a x _a template image of a target image layer C _'a pattern matching, the x _a positive integer.

With reference to the second possible implementation manner of the second aspect of the embodiment of the present invention, in a sixth possible implementation manner of the second aspect of the present invention, when k<3, x _k+1 = 2 ^k ;

The pattern matching module is further configured to the full angle [-π, π] of the target image k + 1 th layer is a template image for pattern matching according to step 2 ^k k + 1 using the first layer.

A third aspect of the embodiments of the present invention provides an apparatus for image pattern matching, including:

a processor, a memory, and a bus, the processor and the memory being connected by a bus;

The memory is used to store a program;

The processor is operative to execute a program in the memory such that the image pattern matching device performs the method of image pattern matching in the first aspect of the invention.

A fourth aspect of the embodiments of the present invention further provides a storage medium. The technical solution of the present invention, which is essential or contributes to the prior art, or all or part of the technical solution may be embodied in the form of a software production port. The computer software product is stored in a storage medium for storing computer software instructions for use in the electronic device, including programs for performing the first, second, and third aspects described above. The computer software product is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and the like. .

It can be seen from the above technical solutions that the embodiments of the present invention have the following advantages:

In the embodiment of the present invention, the original template image and the original target image are first acquired; then the template image pyramid and the target image pyramid of the k+1 layer are established, and the full angle is used at the top layer of the image pyramid. Pattern matching, on the image pyramid of each layer, the interval angle pattern matching, under the premise of ensuring the pattern matching accuracy, the main calculation amount of pattern matching is concentrated on the image of the top layer of the pyramid with the smallest resolution, thereby reducing the total The computational complexity of angle pattern matching improves operational efficiency.

DRAWINGS

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments and the prior art description will be briefly described below. Obviously, the drawings in the following description are only some implementations of the present invention. For example, other drawings may be obtained from those skilled in the art without any inventive effort.

1 is a schematic diagram of an embodiment of a method for image pattern matching according to an embodiment of the present invention;

2 is a schematic diagram of establishing an image pyramid in an embodiment of the present invention;

FIG. 3 is a schematic diagram of another embodiment of a method for image pattern matching according to an embodiment of the present invention; FIG.

4 is a schematic diagram of an embodiment of an image pattern matching apparatus according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of another embodiment of an image pattern matching apparatus according to an embodiment of the present invention; FIG.

FIG. 6 is a schematic diagram of another embodiment of an image pattern matching apparatus according to an embodiment of the present invention.

detailed description

Embodiments of the present invention provide a method for image pattern matching, which is used to reduce the duration of image matching in a pattern matching algorithm.

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. It is an embodiment of the invention, but not all of the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of the present invention.

The terms "first", "second", "third", "fourth", etc. (if present) in the specification and claims of the present invention and the above figures are used to distinguish similar objects without having to use To describe a specific order or order. It is to be understood that the data so used may be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than what is illustrated or described herein. In addition, the terms "comprises" and "comprises" and "comprises", and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited Those steps or units listed are included, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

The following is specifically described by using an embodiment. Referring to FIG. 1 , an embodiment of a method for image pattern matching according to the present invention includes:

101. Obtain an original template image and an original target image;

In this embodiment, the original template image and the original target image can be obtained by the sensor. The original template image and the original target image have various shapes, which are not limited. Generally, the size of the original template image is smaller than the original. The size of the target image, the original template image and the original target image are usually rectangular, for ease of expression, assuming that both the original template image and the original target image are square images.

102. Create a k+1 layer template image pyramid and a target image pyramid. The template image pyramid is reduced from the lowest layer to the top layer, and the area is scaled down by layer. The bottom layer area of the template image pyramid is the area of the template image, and the target image pyramid is from The bottom layer to the top layer, the area is reduced by layer by layer, the bottom layer area of the target image pyramid is the area of the target image, and k is an integer greater than or equal to 1;

In this embodiment, after the original template image and the original target image are acquired, the original template image and the original target image are subjected to pattern matching of normalized cross-correlation. Firstly, the template image pyramid and the target image pyramid of the k+1 layer are established. The template image pyramid is reduced from the lowest layer to the top layer, and the area is scaled down by layer. The bottom layer area of the template image pyramid is the area of the template image, and the target image pyramid is the most From the bottom layer to the top layer, the area is scaled down by layer. The bottom layer area of the target image pyramid is the area of the target image, and k is an integer greater than or equal to 1.

In the prior art, the pattern matching algorithm can calculate a normalized cross-correlation between a template image of size N×N and a target image of size M×M in O(M ² log ₂ M) time. The value (assuming M is greater than N), in the case of higher resolution (M ~ 1000), it usually takes a few seconds to calculate the normalized cross-correlation value; this efficiency is too long for most machine vision applications, and Using image pyramids for fast pattern matching can shorten a certain amount of time.

In general, the size of the original template image is smaller than the size of the original target image, so the height of the image pyramid is determined by the size of the original template image, which is a series of down sampling of a pair of images. For the collection of the image pyramid, please refer to the schematic diagram of the image pyramid of Figure 2, which is straightforward to say the size of the original image and the collection of images reduced to 1/2, 1/4, 1/8 of the original image size, assuming K+1 layer pyramid, then the image pyramid from the lowest to the top, each layer of image area is about a quarter of the previous layer, M × M,

......

In the process of reducing the image, we need to sample. There are generally two sampling methods: Gaussian sampling and Laplacian sampling, which correspond to Gaussian pyramid and Laplacian pyramid respectively. The difference between the two sampling methods is the use of low pass. (Gaussian) filtering also uses bandpass (Laplace) filtering. Compared to bandpass filtering, the implementation of low-pass filtering is simpler, and the filtering used has no significant effect on the application of template matching. Here we use Gaussian pyramids.

Suppose that according to the size of the original template image, a template image pyramid and a target image pyramid of k+1=5 layers are established, and the template image pyramid is from the first layer to the fifth layer, and each layer has an area of four quarters of the upper layer area. First, the first layer area of the template image pyramid is the area of the template image, the target image pyramid is from the first layer to the fifth layer, each layer is one quarter of the area of the upper layer, and the first layer area of the target image pyramid Is the area of the target image.

103. Perform pattern matching on the template image of each layer and the target image of the corresponding layer, and obtain a normalized cross-correlation graph C and a corresponding angle graph A with a normalized cross-correlation value greater than a preset threshold; wherein, the topmost layer The pattern matching between the template image and the topmost target image is matched to the full angle mode, and the pattern images of the other layer's template image and the corresponding layer's target image are matched to the interval angle pattern; the topmost target image is the original target image. At the top level, the target images of the other layers are all sampled from the normalized cross-correlation map acquired by the pattern matching to the next layer.

In this embodiment, after the template image pyramid and the target image pyramid of the k+1 layer are established, the template image of each layer is pattern-matched with the target image of each layer, wherein the template image of each layer and the corresponding layer are Performing pattern matching on the target image, obtaining a normalized cross-correlation map C and a corresponding angle map A whose normalized cross-correlation values are greater than a preset threshold; wherein the pattern of the topmost template image and the topmost target image is matched as Full-angle mode matching, the pattern matching of the template image of the other layer and the target image of the corresponding layer is matched by the interval angle pattern; the top-level target image is obtained when the original target image is reduced to the top layer, and the target image of the other layer is obtained. The normalized cross-correlation map obtained by the pattern matching of the previous layer is sampled to the next layer.

(1) At the top layer, that is, the k+1th layer, pattern matching may include: matching the template image of the k+1th layer with the target image of the k+1th layer to obtain a normalized mutual The normalized cross-correlation graph C _k+1 whose correlation value is greater than the preset threshold and the corresponding angle graph A _k+1 , and the normalized cross-correlation graph C _k+1 and the corresponding angle graph A _k+1 are upsampled k to k-th layer is a layer of target image C _'k and a corresponding angle of FIG. a' _k.

Specifically, the full angle may be [-π, π] or [0, 2π], and is not limited.

Then, performing the full-angle mode matching on the template image of the k+1th layer and the target image of the k+1th layer may include: using the k+1 in the step size x _k+1 in the full angle [-π, π] The template image of the layer performs pattern matching on the target image of the k+1th layer, and x _k+1 is a positive integer.

(2) Performing pattern matching on each of the other layers may include: matching the template image of the layer a and the target image C' _a of the layer a to the interval angle pattern of the layer a, and obtaining the normalized cross-correlation value greater than the preset threshold. The normalized cross-correlation map _Ca and the corresponding angle map A _a , and the normalized cross-correlation map _Ca and the corresponding angle map A _a are upsampled to the target image of the a-1 layer of the a-1 layer C' _a-1 and the corresponding angle diagram A' _a-1 , 1 < a ≤ k, and a is a positive integer.

Specifically, before pattern matching is performed on each of the other layers, it also includes:

a, OK 'within _a communication in each subset, the maximum normalized cross-correlation value corresponding to the position of the A' each other C object image on an angle _a, as

b, according to

Determine the interval angle of the a layer

(n _a is a positive integer).

Then, the template image of the layer a and the target image C' _a of the layer a are matched to the interval angle pattern of the layer a, which may include:

According to the steps of using the first layer a x _a template image of a target image layer C _'a pattern matching, x _a is a positive integer.

The following description is made with an exemplary content, if a 5-layer (k=4) template image pyramid and a target image pyramid are created.

At the fifth level, the top level:

Assuming that the step size x ₅ is 8, it should be noted that the size of the step size x _k+1 is not limited, and the larger the step size is, the smaller the workload is for the machine vision application, and the obtained value is relatively small. Accurate, the smaller the step size, the greater the workload for machine vision applications and the more accurate the values obtained.

Perform pattern matching on the target image of the fifth layer using the template image of the fifth layer according to the step size 8 in the full angle [-π, π], and obtain the normalized cross-correlation graph with the normalized cross-correlation value larger than the preset threshold. C ₅ and the corresponding angle map A ₅ , the normalized cross-correlation map C ₅ and the corresponding angle map A ₅ are upsampled to the fourth layer of the four-layer target image C′ ₄ and the corresponding angle map A′ ₄ , wherein The target image of the fifth layer is obtained when the original target image is reduced to the fifth layer.

Specifically, at the top of the image pyramid, using the angle search based on the rotation template, when the full-angle mode is matched, the angle of the template image rotation is 360 degrees, and the template image of the fifth layer is rotated by -180 degrees, -172 degrees... ...172 degrees, a total of 44 possible angles, respectively, and the 5th layer of the target image for pattern matching, each rotation angle of the template image and the target image will obtain a normalized cross-correlation map, where the normalized cross-correlation The value of the pixel on the graph is [-1,1], and the pixel on the normalized cross-correlation graph takes the normalized cross-correlation value, then 44 normalized cross-correlations of the same size are obtained. According to the 44 normalized cross-correlation maps, the largest normalized cross-correlation value at each position and its corresponding rotation angle are calculated, and a combined normalized cross-correlation map and corresponding angles are obtained. Figure; threshold processing of the integrated normalized cross-correlation map and angle map. Assume that if the preset threshold is set to 0.9, the region larger than 0.9 in the normalized cross-correlation graph is retained while retaining the same region in the angle map. The areas reserved in the two figures constitute a series of connected areas, denoted as C ₅ and A _{5 respectively} . The upsampling mapping is performed on C ₅ and A ₅ to obtain the target image C' ₄ and the corresponding angle map A' ₄ .

At the 4th floor:

a. After obtaining the target image C' ₄ and the corresponding angle map A' ₄ after the C ₅ and A ₅ upsampling mapping, it is also necessary to determine the normalized cross-correlation value in each connected subset of the target image C' ₄ The largest position corresponds to the angle on A' ₄

Here C ', there are ₄ m ₄ interconnecting subsets.

It is known from the above that the acquired target image is C' ₄ and the corresponding angle map A' ₄ , assuming that the maximum position of the cross-correlation value in the first connected subset in the target image C' ₄ corresponds to the template on A' ₄ The angle of rotation of the image is 28 degrees, then

degree. It should be noted that only a connected subset is used for description, and multiple connected subsets are also applicable. The same method is used, and the details of the connected subset are not limited herein.

b, because

The normalized cross-correlation value is the largest in the corresponding connected subset, indicating the corresponding template image rotation

The probability of appearing in the corresponding position in the target image is greater. In the fourth layer, for each connected subset, it can be

Around the angle, that is, about

The interval angle is rotated and searched again. The probability of image pattern matching is higher. It is assumed that n ₄ is 8. In general, the value of n ₄ here will not be too large, because it has already passed once at the top level. Angle screening, the most likely rotation angle of the template image in the first connected subset has been selected

in

Further search around, so the angle of the interval obtained is

c. It is known from the above that the interval angle of the first connected subset of the fourth layer is [20, 36], and the step size x ₄ is assumed to be four. Angle in the interval [20, 36] according to the steps of the template image using the fourth layer 4 of the fourth layer image of the target region corresponding C _'4 communicates a first subset pattern matching; if another communication sub Set, perform similar calculations on other connected subsets, and finally obtain a normalized cross-correlation graph C ₄ with a normalized cross-correlation value greater than a preset threshold and a corresponding angle in combination with the normalized cross-correlation calculation results of all connected subsets. Figure A ₄ , normalized cross-correlation graph C ₄ and angle graph A ₄ upsampled to the third layer of the 3-layer target image C' ₃ and the corresponding angle map A' ₃ . Re-determination target image C ₃ each connected subset 'in the position of maximum normalized cross-correlation value corresponding to the angle

according to

Determine the angle of the third layer

For the first connected subset in C' ₄ , use the angle search based on the rotated template image to rotate the template image of the fourth layer

That is, there are 5 possible angles of 20 degrees, 24 degrees, 28 degrees, 32 degrees, and 36 degrees, respectively, and the target image of the 4th layer corresponds to the area of the C' ₄ first connected subset for pattern matching, and each rotation angle Both the template image and the target image C' ₄ will obtain a normalized cross-correlation map, where the normalized cross-correlation map has a value range of [-1, 1], then 5 normalized sizes of the same size will be obtained. According to the five normalized cross-correlation maps, the largest normalized cross-correlation value at each position and its corresponding rotation angle are calculated, and a comprehensive normalized cross-correlation map is obtained. Corresponding angle map; similar calculations are performed for each connected subset, and all the results are combined to obtain a normalized cross-correlation map and a corresponding angle map. It is known from the above that the preset threshold value is 0.9, and the normalized cross-correlation graph C ₄ corresponding to the normalized cross-correlation value greater than 0.9 and the corresponding angle graph A _{4 are determined} , wherein the normalized cross-correlation graph C ₄ And a series of connected regions of the angle map A ₄ , the upsampling mapping of C ₄ and A _{4 is} performed to obtain the target image C' ₃ and the corresponding angle map A' ₃ .

At the third level:

Suppose, a target image C _'3 in the position of maximum normalized cross-correlation value of the angle corresponding to a first subset of communication

32 degrees, the first connected subset interval angle of the third layer is

Here, assuming n ₃ is 4, then

From the above, it is known that the interval angle of the first connected subset of the third layer is [28, 36], and the step size x ₃ is assumed to be 2. 2 is used in [28, 36] in steps of a template image within the layer 3 of the object image corresponding to the third layer region C _'3 is a first subset of communication pattern matching; if another subset of communication, Perform similar calculations on other connected subsets, and finally obtain a normalized cross-correlation graph C ₃ with a normalized cross-correlation value greater than a preset threshold and a corresponding angle graph A in combination with the normalized cross-correlation calculation results of all connected subsets. _3, normalized cross-correlation C ₃ and FIG angle a _3-sampled FIG mapped to the second layer as a target image C _'2 and FIG angle corresponding a' _{2 2} layers. Re-determination target image C ₂ each connected subset 'in normalized angular position of the maximum cross-correlation value corresponding to

according to

Determine the interval angle of the second layer

For the first connected subset in C' ₃ , use the angle search based on the rotated template image to rotate the template image of the third layer

That is, 28 degrees, 30 degrees, 32 degrees, 34 degrees, 36 degrees, a total of 5 possible angles, respectively, and the target image of the 3rd layer corresponds to the area of the C' ₃ first connected subset for pattern matching, each rotation angle Both the template image and the target image will obtain a normalized cross-correlation graph, where the normalized cross-correlation graph has a value range of [-1, 1], then 5 normalized cross-correlations of the same size will be obtained. According to the five normalized cross-correlation maps, the largest normalized cross-correlation value at each position and its corresponding rotation angle are calculated, and a comprehensive normalized cross-correlation map and corresponding angles are obtained. FIG; then similarly calculated for C _'3 each subset of communication, all the results obtained in conjunction with the normalized cross-correlation and corresponds to the angle of FIG. FIG. It is known from the above that the preset threshold value is 0.9, and the normalized cross-correlation graph C ₃ corresponding to the normalized cross-correlation value greater than 0.9 and the corresponding angle graph A _{3 are determined} , wherein the normalized cross-correlation graph C ₃ And A ₃ is a series of connected regions, and C ₃ and A ₃ are upsampled to obtain a target image C' ₂ and a corresponding angle map A' ₂ .

At the second level:

Assume that the angle corresponding to the maximum position of the normalized cross-correlation value in the first connected subset of the target image C' ₂

At 30 degrees, the interval angle of the first connected subset of the second layer is

Here, assuming n ₂ is 2, then

From the above, it is known that the interval angle of the first connected subset of the second layer is [28, 32], and the step size x ₂ is assumed to be 1. Perform pattern matching on the region of the second layer whose target image corresponds to the first connected subset of C' ₂ in the [28, 32] using the template image of the second layer in steps of 1; if there are other connected subsets, Perform similar calculations on other connected subsets, and finally obtain a normalized cross-correlation graph C ₂ with a normalized cross-correlation value greater than a preset threshold and a corresponding angle graph A in combination with the normalized cross-correlation calculation results of all connected subsets. _{2. The} normalized cross-correlation graph C ₂ and the angle map A ₂ are upsampled to the target image C' ₁ of the first layer and the corresponding angle map A' ₁ . Re-determination target image C ₁ of each connected subset 'in normalized cross-correlation value is the largest angle corresponding to the position

according to

Determine the angle of the first layer

For the first connected subset in C' ₂ , use the angle search based on the rotated template image to rotate the template image of the second layer

That is, there are 5 possible angles of 28 degrees, 29 degrees, 30 degrees, 31 degrees, and 32 degrees, and pattern matching is performed respectively with the target image of the second layer corresponding to the area of the C' ₂ first connected subset, and each rotation angle Both the template image and the target image will obtain a normalized cross-correlation graph, where the normalized cross-correlation graph has a value range of [-1, 1], then 5 normalized cross-correlations of the same size will be obtained. According to the five normalized cross-correlation maps, the largest normalized cross-correlation value at each position and its corresponding rotation angle are calculated, and a comprehensive normalized cross-correlation map and corresponding angles are obtained. A similar calculation is performed for each connected subset, and all the results are combined to obtain a normalized cross-correlation map and a corresponding angle map. It is known from the above that the preset threshold value is 0.9, and the normalized cross-correlation graph C ₂ corresponding to the normalized cross-correlation value greater than 0.9 and the corresponding angle graph A _{2 are determined} , wherein the normalized cross-correlation graph C ₂ And A ₂ is a series of connected regions, and C ₂ and A ₂ are upsampled to obtain a target image C' ₁ and a corresponding angle map A' ₁ .

104, when the first layer, obtaining a normalized cross-correlation value greater than a preset threshold value of the normalized cross-correlation FIG C ₁ represents the original target image of the original template image and, C corresponding to the area A _{1 1} corresponding to the angle of FIG. The position appearing in the corresponding angle angle corresponding to the angle map A ₁ represents the rotation angle of the original template image at the position where the original target image appears.

In this embodiment, when the first layer is obtained, the normalized cross-correlation graph C ₁ with the normalized cross-correlation value greater than the preset threshold is obtained, and the corresponding angle map A ₁ , the area corresponding to C ₁ represents the original template image. The position appearing in the original target image, the angle corresponding to the corresponding angle map A ₁ represents the rotation angle of the original template image at the position where the original target image appears.

Exemplary, when the first layer is assumed that the target image C ', _a first subset of the communication in the normalized cross-correlation maximum normalized cross-correlation values corresponding to the angle of FIG.

At 29 degrees, the interval angle of the first connected subset of the first layer is

Here, assuming n ₁ is 1, then

From the above, it is known that the interval angle of the first connected subset of the second layer is [28, 30], and the step size x ₁ is assumed to be 1. In [28, 30] according to the steps of a template image used for the first layer of the object image of the first layer corresponding to the region C _'1 of the first subset of communication pattern matching; for the remaining subset of communication Similar to the calculation, the normalized cross-correlation graph C ₁ with the normalized cross-correlation value greater than the preset threshold and the corresponding angle graph A _{1 are} finally obtained by combining the normalized cross-correlation calculation results of all connected subsets.

For C ₁ a first subset of communication ', using the rotation angle of the search based on the template image, the template image of the first layer of rotation

I.e., 28 degrees, 29 degrees, 30 degrees, a total of three possible angles, respectively, and the object image of the first layer corresponding to the region C _'1 of the first subset of communication pattern matching, the template image and the target image for each rotation angle A normalized cross-correlation graph is obtained, wherein the normalized cross-correlation graph has a value range of [-1, 1], then three normalized cross-correlation graphs of the same size are obtained, according to the three Normalize the cross-correlation map, calculate the largest normalized cross-correlation value at each position and its corresponding rotation angle, and obtain a combined normalized cross-correlation map and corresponding angle map; The connected subsets are similarly calculated, and all the results are combined to obtain a normalized cross-correlation map and corresponding angle maps. It is known from the above that the preset threshold value is 0.9, and the normalized cross-correlation graph C ₁ corresponding to the normalized cross-correlation value greater than 0.9 and the corresponding angle graph A _{1 are determined} , wherein the normalized cross-correlation graph C ₁ And A ₁ is a series of connected areas.

Therefore, in the first layer, the normalized cross-correlation graph C ₁ with the normalized cross-correlation value greater than the preset threshold is obtained, and the normalized cross-correlation value of each connected subset in the corresponding angle graph A ₁ , C ₁ is obtained. The largest position represents the most likely position of the original template image in the original target image, and the corresponding angle of the corresponding angle map A ₁ represents the most likely rotation angle of the original template image when the original target image appears. .

It should be noted that, in the above example, the normalized cross-correlation graphs C' ₁ , C' ₂ , C' ₃ , C' ₄ which are larger than the preset threshold in each layer may actually include a plurality of connected components. Set, the above is only used to illustrate one of them, that is, the first connected subset. Similarly, if there are multiple connected subsets, the same calculation method is used to determine the interval angle.

In the embodiment of the present invention, the full-angle pattern matching is used at the top of the image pyramid, and the interval angle pattern matching is performed on the non-topmost image pyramid. The calculation amount of the pattern matching is ensured by the present invention under the premise of ensuring the pattern matching precision. Focusing on low-resolution pyramid images reduces the computational complexity of full-angle pattern matching and improves operational efficiency. Because of the actual application, pattern matching usually requires full-angle search because the invention is real-time and efficient for development. The machine vision application will be of great help.

Referring to FIG. 3, another embodiment of a method for image pattern matching according to the present invention includes:

301. Obtain an original template image and an original target image;

In the embodiment of the present invention, step 301 is the same as step 101 shown in FIG. 1, and details are not described herein again.

302. Establish a 3-layer template image pyramid and a target image pyramid. The template image pyramid is reduced from the lowest layer to the top layer, and the area is reduced by layer by layer. The bottom layer area of the template image pyramid is original. The area of the template image, the target image pyramid from the lowest to the top, the area is reduced by layer by layer, the bottom layer area of the target image pyramid is the area of the original target image, and k is an integer greater than or equal to 1;

In the embodiment of the present invention, after acquiring the original template image and the original target image, a 3-layer template image pyramid and a target image pyramid are established, indicating that the original template image has a relatively small area. The template image pyramid is from the bottom to the top, and the area is scaled down by layer. The bottom layer area of the template image pyramid is the area of the original template image; the target image pyramid is from the bottom to the top, and the area is scaled down by layer. The bottom layer area of the image pyramid is the area of the original target image, and k is an integer greater than or equal to 1.

303, when the third layer, the template image and the target image layer, the third layer, the third full angle of pattern matching, obtaining a normalized cross-correlation value greater than a preset threshold value of the normalized cross-correlation FIGS C ₃ and the corresponding An angle view A ₃ , and upsample the normalized cross-correlation map C ₃ and the corresponding angle map A ₃ to the second layer of the target image C' ₂ of the second layer and the corresponding angle map A'₂;

In the embodiment of the present invention, in the third layer, the template image of the third layer is matched with the target image of the third layer by full-angle mode, and the normalized cross-correlation whose normalized cross-correlation value is greater than the preset threshold is obtained. C ₃ and FIG angles corresponding to FIG. a _3, and normalized cross-correlation C ₃ and FIG angle corresponding to FIG sample layer 2 second layer is a target image C _'2 and FIG angle corresponding a' on the a _{3 2} .

Specifically, the full angle is [-π, π], which may include: using the template image of the k+1th layer to the target of the k+1th layer according to the step size x _k+1 in the full angle [-π, π] The image is pattern matched and x _k+1 is a positive integer. It should be noted that since the established image pyramid is a pyramid with a k+1 layer equal to 3, when k<3, x _k+1 = 2 ^k .

Then, in the third layer, x ₃ = 2 ² = 4, in the full angle [-π, π] according to the step size 4 using the template image of the third layer to perform pattern matching on the target image of the third layer, obtaining the return Normalized cross-correlation graph C ₃ and corresponding angle graph A ₃ having a cross-correlation value greater than a preset threshold, and up-sampling the normalized cross-correlation graph C ₃ and the corresponding angle graph A ₃ to the second layer The 2-layer target image C' ₂ and the corresponding angle map A' ₂ .

304. In the second layer, the template image of the second layer and the target image C' ₂ of the second layer are matched by the interval angle pattern of the two layers, and the normalized cross-correlation value is greater than the preset threshold. Correlating the graph C ₂ and the corresponding angle map A ₂ , and upsampling the normalized cross-correlation graph C ₂ and the corresponding angle graph A ₂ to the target image C′ ₁ of the first layer and the corresponding angle graph A '₁;

In the embodiment of the present invention, in the second layer, the template image of the second layer and the target image C' ₂ of the second layer are matched by the two-layer interval angle pattern, and the normalized cross-correlation value is greater than the preset threshold. Normalized cross-correlation graph C ₂ and corresponding angle graph A ₂ , and upsampled normalized cross-correlation graph C ₂ and corresponding angle graph A ₂ to target image C' _{1 of} layer 1 of layer ₁ and Corresponding angle diagram A' ₁ .

It should be noted that, before the interval angle mode matching is performed, the method further includes: determining, in each connected subset of the target image C′ ₂ of the second layer, the position where the normalized cross-correlation value is the largest corresponds to the A′ ₂ Angle, for

according to

Determine the interval angle of the 2 layers

(n ₂ is a positive integer). Here n ₂ can be 2 ¹ = 2.

Specifically, in the second layer, at the interval angle

Performing pattern matching on the target image C' _{2 of} the second layer using the template image of the second layer according to the step size 2, and obtaining the normalized cross-correlation graph C ₂ and the corresponding normalized cross-correlation value greater than the preset threshold. The angle map A ₂ and the normalized cross-correlation map C ₂ and the corresponding angle map A ₂ are upsampled to the target image C' ₁ of the first layer and the corresponding angle map A' ₁ .

305, when the first layer, obtaining a normalized cross-correlation value greater than a preset threshold value of the normalized cross-correlation FIG C ₁ represents the original target image of the original template image and, C corresponding to the area A _{1 1} corresponding to the angle of FIG. The position appearing in the corresponding angle angle corresponding to the angle map A ₁ represents the rotation angle of the original template image at the position where the original target image appears.

In the embodiment of the present invention, when the first layer is obtained, the normalized cross-correlation graph C ₁ with the normalized cross-correlation value greater than the preset threshold is obtained, and the corresponding angle map A ₁ , the region corresponding to C ₁ represents the original template image. At the position appearing in the original target image, the angle corresponding to the corresponding angle map A ₁ represents the rotation angle of the original template image at the position where the original target image appears.

It should be noted that, before obtaining the normalized cross-correlation graph C ₁ and the corresponding angle graph A ₁ whose normalized cross-correlation value is greater than a preset threshold, the method further includes: determining each target image C′ ₁ of the first layer Within the connected subset, the position with the largest normalized cross-correlation value corresponds to the angle on A' ₁

according to

Determine the interval angle of the 1st floor

(n ₁ is a positive integer). Here n ₁ can be 2 ⁰ =1.

Specifically, at the first layer, at the interval angle

Performing pattern matching on the target image C' ₁ of the first layer using the template image of the first layer according to the step size of 1 to obtain a normalized cross-correlation graph C ₁ with a normalized cross-correlation value greater than a preset threshold and corresponding Angle view A ₁ . Then, the area corresponding to C ₁ represents the position where the original template image appears in the original target image, and the angle corresponding to the corresponding angle map A ₁ represents the rotation angle of the original template image at the position where the original target image appears.

It should be noted that, in the embodiment of the present invention, the specific pattern matching process is not described in detail herein, and the content expressed in FIG. 1 above may be referred to.

The method for image pattern matching is described above, and the method is applied to the image pattern matching device. The device is described below. Referring to FIG. 4, an embodiment of the image pattern matching device provided by the present invention includes:

a first obtaining module 401, configured to acquire an original template image and an original target image;

The module 402 is configured to establish a template image pyramid and a target image pyramid of the k+1 layer. The original template image pyramid is reduced from the lowest layer to the top layer, and the area is reduced by layer by layer. The bottom layer area of the template image pyramid is original. The area of the template image, the target image pyramid from the lowest to the top, the area is reduced by layer by layer, the bottom layer area of the target image pyramid is the area of the original target image, and k is an integer greater than or equal to 1;

The pattern matching module 403 is configured to perform pattern matching between the template image of each layer and the target image of the corresponding layer, and obtain a normalized cross-correlation graph C and a corresponding angle map A with normalized cross-correlation values greater than a preset threshold; Wherein, the pattern of the topmost template image matches the pattern of the topmost target image is a full-angle pattern matching, and the pattern images of the other layer's template image and the corresponding layer's target image are matched to the interval angle pattern; the topmost target image is original. When the target image is reduced to the top layer, the target images of the other layers are all sampled by the previous layer for pattern matching, and the normalized cross-correlation map is sampled to the next layer;

A second acquiring module 404, a first layer, obtaining a normalized cross-correlation value greater than a preset threshold value, a normalized cross-correlation C of FIG. ₁ and represents the original template, C corresponding to the region corresponding to the angle A _{1 1} FIG. The position at which the image appears in the original target image, and the angle corresponding to the corresponding angle map A ₁ represents the rotation angle of the original template image at the position where the original target image appears.

Optionally, in some embodiments of the invention,

The pattern matching module 403 is specifically configured to perform a full-angle pattern matching between the template image of the k+1th layer and the target image of the k+1th layer, and obtain a normalized cross-correlation graph whose normalized cross-correlation value is greater than a preset threshold. C _k+1 and the corresponding angle map A _k+1 , and upsample the normalized cross-correlation map C _k+1 and the corresponding angle map A _k+1 to the target image C' _k of the k-th layer And the corresponding angle map A' _k .

Optionally, in some embodiments of the invention, the full angle is [-π, π],

The pattern matching module 403 is further configured to perform pattern matching on the target image of the k+1th layer by using the template image of the k+1th layer according to the step size x _k+1 in the full angle [-π, π], x _{k +1} is a positive integer.

Optionally, in some embodiments of the invention,

The pattern matching module 403 is further configured to perform the interval angle pattern matching of the layer image of the layer a and the target image C′ _a of the layer a to obtain a normalized cross-correlation value greater than a preset threshold. The cross-correlation map _Ca and the corresponding angle map A _a , and up-sampling the normalized cross-correlation map _Ca and the corresponding angle map A _a to the target image C' _{a- of} the a-1 layer of the a-1st layer ₁ and the corresponding angle diagram A' _a-1 , 1 < a ≤ k, and a is a positive integer.

Optionally, in some embodiments of the present invention, as shown in FIG. 4, as shown in FIG. 5, the apparatus further includes:

First determining module 405 for determining 'of the communication of each of _a subset of the normalized cross-correlation maximum values corresponding to the position A' of each other C object image on an angle _a, as

a second determining module 406, configured to

Determine the interval angle of the a layer

(n _a is a positive integer).

Optionally, in some embodiments of the invention,

The pattern matching module 403 is specifically used for the interval angle

According to the steps of using the first layer a x _a template image of a target image layer C _'a pattern matching, x _a is a positive integer.

Optionally, in some embodiments of the present invention, when k<3, x _k+1 = 2 ^k ;

Pattern matching module 403, is also used to specifically target image k + 1 th layer is a template image for pattern matching according to step 2 ^k k + 1 using the first layer in the whole angle [-π, π] within.

As shown in FIG. 6, another embodiment of an apparatus for image pattern matching in an embodiment of the present invention includes:

The memory 601, the processor 602 and the bus 603; the memory 601 and the processor 602 are connected by a bus 603; the memory 601 is configured to store application code for executing the method executed by the media stream transmitting device in the solution of the present invention, and is controlled by the processor 602. carried out. The processor 602 is configured to execute application code stored in the memory.

The memory 601 can be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type that can store information and instructions. The dynamic storage device can also be an Electrically Erasable Programmable Read-Only Memory (EEPROM) or a Compact Disc Read-Only Memory (CD-ROM). Or other disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), disk storage media or other magnetic storage devices, or can be used to carry or store expectations in the form of instructions or data structures Program code and any other medium that can be accessed by a computer, but is not limited thereto. The memory can exist independently and be connected to the processor via a bus. The memory can also be integrated with the processor.

Processor 602 can be a general purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the present invention. It can also be an integrated circuit chip with signal processing capability, which can be a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device. Each of the methods, steps, and logic blocks in the embodiments of the present invention may be implemented or executed by a discrete gate or a transistor logic device or a discrete hardware component.

Bus 603 can include a path for communicating information between the components described above.

Specifically, the processor is used to perform the steps in the foregoing method for image pattern matching in FIG. 1 or FIG. 3, and details are not described herein again.

The embodiment of the invention further provides a computer storage medium for storing computer software instructions for the image pattern matching device of FIG. 4 or FIG. 5, which comprises a program designed to execute the above method embodiment. By executing a stored program, the length of image matching in the pattern matching algorithm can be reduced.

A person skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the system, the device and the unit described above can refer to the corresponding process in the foregoing method embodiment, and details are not described herein again.

In the several embodiments provided by the present application, it should be understood that the disclosed system, apparatus, and method may be implemented in other manners. For example, the device embodiments described above are merely illustrative. For example, the division of the unit is only a logical function division. In actual implementation, there may be another division manner, for example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be in an electrical, mechanical or other form.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a standalone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention, which is essential or contributes to the prior art, or all or part of the technical solution, may be embodied in the form of a software product stored in a storage medium. A number of instructions are included to cause a computer device (which may be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and the like. .

The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to be limiting; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art will understand that The technical solutions described in the embodiments are modified, or the equivalents of the technical features are replaced by the equivalents of the technical solutions of the embodiments of the present invention.

Claims (15)

1. A method for image pattern matching, comprising:

   Acquiring the original template image with the original target image;

   Establishing a template image pyramid and a target image pyramid of a k+1 layer, the template image pyramid is reduced from the lowest layer to the top layer, and the area is scaled down by layer, and the bottom layer area of the template image pyramid is the original template image. An area, the target image pyramid is reduced from the lowest layer to the top layer, and the area is scaled down by layer. The bottom layer area of the target image pyramid is the area of the original target image, and k is an integer greater than or equal to 1;

   Performing pattern matching on the template image of each layer and the target image of the corresponding layer to obtain a normalized cross-correlation graph C and a corresponding angle map A with normalized cross-correlation values greater than a preset threshold; wherein, the topmost template image The pattern matching with the topmost target image is matched to the full angle pattern, and the pattern images of the other layer's template image and the target image of the corresponding layer are matched to the interval angle pattern; the topmost target image is the original target image, etc. The normalized cross-correlation map obtained by pattern matching obtained from the previous layer is obtained by zooming to the top layer, and the target image of the other layer is sampled to the next layer.

   In the first layer, a normalized cross-correlation map C _{1 having} a normalized cross-correlation value greater than the preset threshold is obtained, and a corresponding angle map A ₁ , where the area corresponding to C ₁ indicates that the original template image is the original position of the target image appears, a ₁ corresponds to the angle corresponding to an angle view showing the rotation angle position of the original template image appearing in the original target image.
2. The method of claim 1 wherein the full-angle pattern matching of the topmost template image with the topmost target image comprises:

   Performing a full-angle pattern matching between the template image of the k+1th layer and the target image of the k+1th layer, and obtaining a normalized cross-correlation graph _Ck+1 and corresponding to the normalized cross-correlation value greater than the preset threshold An angle map A _k+1 , and up-samples the normalized cross-correlation map C _k+1 and the corresponding angle map A _k+1 to the target image C′ _k of the k-th layer and the corresponding angle Figure A' _k .
3. The method according to claim 2, wherein said full angle is [-π, π],

   The matching the template image of the k+1th layer with the target image of the k+1th layer is performed in an all-angle mode, including:

   Performing pattern matching on the target image of the k+1th layer using the template image of the k+1th layer in the full angle [-π, π] according to the step size x _k+1 , the x _k+1 being a positive integer .
4. The method according to claim 1, wherein the interval angle pattern matching performed by the template image of each of the other layers and the target image of the corresponding layer comprises:

   Angle interval pattern matching of a template image and the second layer a layer of the target image C _'a a layer for obtaining a normalized cross-correlation value is greater than the preset threshold value of the normalized cross-correlation C _a in FIG. And the corresponding angle map A _a , and up-sampling the normalized cross-correlation map _Ca and the corresponding angle map A _a to the target image C' _a-1 of the a-1 layer and corresponding to the a-1 layer FIG angle _{a 'a-1, 1 <} a≤k, a is a positive integer and.
5. The method of claim 4, wherein the method further comprises:

   Determining, in each connected subset of the target images C' _a of each of the other layers, the position of the normalized cross-correlation value corresponding to the angle on A' _a is

   

   Said

   

   According to the

   

   Determine the interval angle of the a layer

   

   (n _a is a positive integer).
6. Method according to claim 4 or 5, characterized in that

   The matching the template image of the layer a and the target image C' _a of the layer a to the interval angle pattern of the layer a includes:

   In the interval angle

   

   According to the steps of using the first layer a x _a template image of a target image layer C _'a pattern matching, said x _a positive integer.
7. The method according to claim 3, wherein when k < 3, x _k+1 = 2 ^k ;

   Performing pattern matching on the target image of the k+1th layer using the template image of the k+1th layer according to the step size _xk+1 in the full angle [-π, π], including:

   Target image k + 1 th layer is a template image for pattern matching according to step 2 ^k k + 1 using the first layer in said full angle [-π, π] within.
8. An apparatus for image pattern matching, comprising:

   a first acquiring module, configured to acquire an original template image and an original target image;

   A module is established for establishing a template image pyramid and a target image pyramid of the k+1 layer, the original template image pyramid is reduced from the lowest layer to the top layer, and the area is reduced by layer by layer, and the bottom layer area of the template image pyramid is The area of the original template image, the target image pyramid is reduced from the lowest layer to the top layer, and the area is scaled down by layer. The bottom layer area of the target image pyramid is the area of the original target image, and the k is An integer greater than or equal to 1;

   The pattern matching module is configured to pattern match the template image of each layer with the target image of the corresponding layer, and obtain a normalized cross-correlation graph C and a corresponding angle map with the normalized cross-correlation value greater than a preset threshold. A; wherein the pattern of the topmost template image matches the pattern of the topmost target image is a full angle pattern matching, and the pattern image of the template image of the other layer matches the pattern of the target image of the corresponding layer to the interval angle pattern matching; the topmost target The image is obtained when the original target image is reduced to the topmost level, and the target images of the other layers are all sampled by the previous layer for pattern matching, and the normalized cross-correlation map is sampled to the next layer;

   a second obtaining module, when used in the first layer, obtains a normalized cross-correlation graph C ₁ and a corresponding angle graph A ₁ whose normalized cross-correlation value is greater than the preset threshold, and the area corresponding to the C ₁ a position at which the original template image appears in the original target image, and an angle corresponding to the corresponding angle map A ₁ represents a rotation of the original template image at a position appearing in the original target image angle.
9. The device of claim 8 wherein:

   The mode matching module is configured to perform a full-angle mode matching between the template image of the k+1th layer and the target image of the k+1th layer, and obtain a normalization that the normalized cross-correlation value is greater than the preset threshold. Correlation graph C _k+1 and corresponding angle graph A _k+1 , and up-sampling the normalized cross-correlation graph C _k+1 and the corresponding angle graph A _k+1 to the k-th layer The target image _C'k and the corresponding angle map _A'k .
10. The apparatus according to claim 9, wherein said full angle is [-π, π],

    The pattern matching module is further configured to perform pattern matching on the target image of the k+1th layer by using the template image of the k+1th layer according to the step size x _k+1 in the full angle [-π, π] the x _{k + 1} is a positive integer.
11. The device of claim 8 wherein:

    The pattern matching module is further configured to layer a first template image with the target image of a layer C _'a layer for a pattern matching angle section, obtaining a normalized cross-correlation value is greater than the preset The normalized cross-correlation map C _{a of the} threshold and the corresponding angle map A _a , and up-sampling the normalized cross-correlation map _Ca and the corresponding angle map A _a to the a-1 layer The target image C'a _-1 and the corresponding angle map A'a _-1 , 1<a≤k, and a is a positive integer.
12. The device according to claim 11, wherein the device further comprises:

    a first determining module, configured to determine, in each connected subset of the target images C' _a of each of the other layers, the position where the normalized cross-correlation value is the largest corresponds to an angle on A' _a ,

    

    Said

    

    a second determining module for

    

    Determine the interval angle of the a layer

    

    (n _a is a positive integer).
13. Device according to claim 11 or 12, characterized in that

    The pattern matching module is specifically used for the interval angle

    

    According to the steps of using the first layer a x _a template image of a target image layer C _'a pattern matching, said x _a positive integer.
14. The apparatus according to claim 10, wherein when k < 3, x _k+1 = 2 ^k ;

    The pattern matching module is further configured to use a template image k + 1 th layer in said full angle [-π, π] 2 ^k according to the steps of the target image k + 1 th layer pattern matching.
15. An apparatus for image pattern matching, comprising:

    Memory, processor and bus;

    The memory and the processor are connected by a bus;

    a memory for storing application code executed by the apparatus, the processor being configured to execute an application stored in the memory;

    The processor is configured to acquire an original template image and an original target image;

    Establishing a template image pyramid and a target image pyramid of a k+1 layer, the template image pyramid is reduced from the lowest layer to the top layer, and the area is scaled down by layer, and the bottom layer area of the template image pyramid is the original template image. An area, the target image pyramid is reduced from the lowest layer to the top layer, and the area is scaled down by layer. The bottom layer area of the target image pyramid is the area of the original target image, and k is an integer greater than or equal to 1;

    Performing pattern matching on the template image of each layer and the target image of the corresponding layer to obtain a normalized cross-correlation graph C and a corresponding angle map A with normalized cross-correlation values greater than a preset threshold; wherein, the topmost template image The pattern matching with the topmost target image is matched to the full angle pattern, and the pattern images of the other layer's template image and the target image of the corresponding layer are matched to the interval angle pattern; the topmost target image is the original target image, etc. The normalized cross-correlation map obtained by pattern matching obtained from the previous layer is obtained by zooming to the top layer, and the target image of the other layer is sampled to the next layer.

    In the first layer, a normalized cross-correlation map C _{1 having} a normalized cross-correlation value greater than the preset threshold is obtained, and a corresponding angle map A ₁ , where the area corresponding to C ₁ indicates that the original template image is the original position of the target image appears, a ₁ corresponds to the angle corresponding to an angle view showing the rotation angle position of the original template image appearing in the original target image.

Images